TheoremBase

The cutoff is a scaling substitution of the bump function, so its partial derivatives are the rescaled partial derivatives of the bump, which are bounded because they are continuous with compact support; the product rule then gives the derivatives of the second-moment test functions, which are estimated on the ball of radius 2R and vanish outside it.

Proof

Each result cited is universally quantified over the data in its own statement. Write ι\iota for the canonical map of The Canonical Map from the Natural Numbers to a Field, πl:Rq→R\pi_{l}:\mathbb{R}^{q}\to\mathbb{R}, πl(x)=xl\pi_{l}(x)=x_{l}, for the coordinate functions, and W2={y∈Rq:∥y∥>2}W_{2}=\{y\in\mathbb{R}^{q}:\lVert y\rVert>2\}. Continuity of a function on Rq\mathbb{R}^{q} is that of Differential Calculus and Convexity on Euclidean Open Sets: Standing Notation (iv); by claim 1 of Euclidean Continuity Agrees with Metric Continuity for Real-Valued Functions it agrees with continuity at every point in the sense of Continuity at a Point for Maps Between Euclidean Spaces, which is the notion used in Functions of Class Cᵏ on an Open Subset of ℝⁿ. A function of class Ck+1C^{k+1} is of class C1C^{1} and its partial derivatives are of class CkC^{k}, by clause 2 of Functions of Class Cᵏ on an Open Subset of ℝⁿ; a smooth function is of class CkC^{k} for every kk (Smooth Functions on an Open Subset of ℝⁿ), so its partial derivatives are smooth as well. A function of class C1C^{1} is continuous with continuous partial derivatives, by clause 1 of Functions of Class Cᵏ on an Open Subset of ℝⁿ.

For points y,y′∈Rqy,y'\in\mathbb{R}^{q} we use y=y′+(y−y′)y=y'+(y-y') and ∥y−y′∥=∥y′−y∥\lVert y-y'\rVert=\lVert y'-y\rVert (claims 2 and 3 of Euclidean Space Rn\mathbb{R}^n is a Real Vector Space, and claim 5 of Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square with λ=−1\lambda=-1), together with dE(y,y′)=∥y−y′∥d_{E}(y,y')=\lVert y-y'\rVert (claim 2 of Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square).

Step 0: the derivatives of χ\chi are bounded and vanish on W2W_{2}. The set W2W_{2} is open: if y∈W2y\in W_{2} and ∥y′−y∥<∥y∥−2\lVert y'-y\rVert<\lVert y\rVert-2, then ∥y∥≤∥y′∥+∥y−y′∥\lVert y\rVert\le\lVert y'\rVert+\lVert y-y'\rVert by claim 6 of Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square applied to y=y′+(y−y′)y=y'+(y-y'), so ∥y′∥>2\lVert y'\rVert>2; this says that the open ball of centre yy and radius ∥y∥−2\lVert y\rVert-2 lies in W2W_{2}, so W2W_{2} is open in the metric sense, hence open by Euclidean Openness Agrees with Metric Openness on Rn\mathbb{R}^n. Let u:Rq→Ru:\mathbb{R}^{q}\to\mathbb{R} be of class C1C^{1} with u=0u=0 on W2W_{2} and let l∈[q]l\in[q]. Then u∣W2=0⋅u∣W2u|_{W_{2}}=0\cdot u|_{W_{2}} is of class C1C^{1} on W2W_{2} by claim 3 of Restriction of a Cᵏ Function to an Open Subset, so by the scalar-multiple rule of claim 1 of Constants, Affine Functions, Sums, Products and Quotients of Cᵏ Functions, ∂l(u∣W2)=0⋅∂l(u∣W2)=0\partial_{l}(u|_{W_{2}})=0\cdot\partial_{l}(u|_{W_{2}})=0 on W2W_{2}, and by claim 1 of Restriction of a Cᵏ Function to an Open Subset, ∂lu=0\partial_{l}u=0 on W2W_{2}. Applying this to u=χu=\chi and then to u=∂iχu=\partial_{i}\chi (which is smooth, hence of class C1C^{1}, and vanishes on W2W_{2} by the first application), we get ∂iχ=0\partial_{i}\chi=0 and ∂j∂iχ=0\partial_{j}\partial_{i}\chi=0 on W2W_{2} for all i,j∈[q]i,j\in[q]. These functions are continuous on Rq\mathbb{R}^{q} (as χ\chi is of class C2C^{2}) and vanish whenever ∥y∥>2\lVert y\rVert>2, so they are compactly supported by claim 2 of Compact Support on Rn\mathbb{R}^n Means Vanishing Outside a Bounded Set and bounded by claim 1 of A Continuous Compactly Supported Function on Rn\mathbb{R}^n is Bounded and Integrable. For i,j∈[q]i,j\in[q] choose bounds BiB_{i} and BjiB_{ji}, nonnegative real numbers by that definition, with ∣∂iχ(y)∣≤Bi|\partial_{i}\chi(y)|\le B_{i} and ∣∂j∂iχ(y)∣≤Bji|\partial_{j}\partial_{i}\chi(y)|\le B_{ji} for every yy, and put

M1=∑i=1qBi,M2=∑i=1q(∑j=1qBji),M_{1}=\sum_{i=1}^{q}B_{i},\qquad M_{2}=\sum_{i=1}^{q}\Bigl(\sum_{j=1}^{q}B_{ji}\Bigr),

finite sums of nonnegative real numbers. By claim 6 of Properties of Finite Sums (each summand of a sum of nonnegative terms is at most the sum, applied once for M1M_{1} and twice for M2M_{2}), Bi≤M1B_{i}\le M_{1} and Bji≤M2B_{ji}\le M_{2} for all i,ji,j, and M1,M2≥0M_{1},M_{2}\ge0 by claim 5 there.

Step 1: claim 1. Fix R>0R>0; R−1>0R^{-1}>0 by claim 7 of Elementary Order Arithmetic in an Ordered Field, so ∣R−1∣=R−1|R^{-1}|=R^{-1} by Absolute Value in an Ordered Field, and claim 5 of Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square gives ∥R−1x∥=R−1∥x∥\lVert R^{-1}x\rVert=R^{-1}\lVert x\rVert for every xx. Multiplying by R−1R^{-1}, respectively by RR, using claim 5 of Elementary Arithmetic in an Ordered Field for the weak inequalities and claim 10 of Elementary Order Arithmetic in an Ordered Field for the strict one, together with R R−1=1R\,R^{-1}=1, we get for every xx:

∥x∥≤R  ⟺  ∥R−1x∥≤1,∥x∥≥2R  ⟺  ∥R−1x∥≥2,∥x∥>2R  ⟺  ∥R−1x∥>2.\lVert x\rVert\le R\iff\lVert R^{-1}x\rVert\le1,\qquad \lVert x\rVert\ge2R\iff\lVert R^{-1}x\rVert\ge2,\qquad \lVert x\rVert>2R\iff\lVert R^{-1}x\rVert>2 .

Consequently 0≤χR≤10\le\chi_{R}\le1, χR(x)=1\chi_{R}(x)=1 whenever ∥x∥≤R\lVert x\rVert\le R, χR(x)=0\chi_{R}(x)=0 whenever ∥x∥≥2R\lVert x\rVert\ge2R, and χR\chi_{R} is compactly supported by claim 2 of Compact Support on Rn\mathbb{R}^n Means Vanishing Outside a Bounded Set.

Apply Partial Derivatives, Continuity and CkC^k Regularity under a Scaling Substitution with n=qn=q, m=1m=1, c=0Rqc=0_{\mathbb{R}^{q}}, λ=R−1\lambda=R^{-1}, μ=1\mu=1, U=RqU=\mathbb{R}^{q} (open by claim 1 of Euclidean Space Is Open in Itself) and f=χf=\chi: since 0Rq+R−1x=R−1x0_{\mathbb{R}^{q}}+R^{-1}x=R^{-1}x (claim 1 of Euclidean Space Rn\mathbb{R}^n is a Real Vector Space), the function gg there is χR\chi_{R} and V=RqV=\mathbb{R}^{q}. By claim 5 there χR\chi_{R} is smooth, and by claim 2 there, for every xx and i∈[q]i\in[q],

∂iχR(x)=R−1 ∂iχ(R−1x).\partial_{i}\chi_{R}(x)=R^{-1}\,\partial_{i}\chi(R^{-1}x).

Apply the same lemma with the same c,λ,μ,Uc,\lambda,\mu,U to f=∂iχf=\partial_{i}\chi (smooth): the function gi(x)=∂iχ(R−1x)g_{i}(x)=\partial_{i}\chi(R^{-1}x) has, by claim 2 there, ∂jgi(x)=R−1∂j∂iχ(R−1x)\partial_{j}g_{i}(x)=R^{-1}\partial_{j}\partial_{i}\chi(R^{-1}x) for every xx and j∈[q]j\in[q]. Since ∂iχR=R−1gi\partial_{i}\chi_{R}=R^{-1}g_{i} on Rq\mathbb{R}^{q}, the scalar-multiple rule of claim 1 of Constants, Affine Functions, Sums, Products and Quotients of Cᵏ Functions gives

∂j∂iχR(x)=R−1 ∂jgi(x)=R−2 ∂j∂iχ(R−1x),\partial_{j}\partial_{i}\chi_{R}(x)=R^{-1}\,\partial_{j}g_{i}(x)=R^{-2}\,\partial_{j}\partial_{i}\chi(R^{-1}x),

where R−2=R−1R−1R^{-2}=R^{-1}R^{-1}. By Step 0 and claim 4 of Properties of the Absolute Value in an Ordered Field, ∣∂iχR(x)∣≤R−1Bi≤M1R−1|\partial_{i}\chi_{R}(x)|\le R^{-1}B_{i}\le M_{1}R^{-1} and ∣∂j∂iχR(x)∣≤R−2Bji≤M2R−2|\partial_{j}\partial_{i}\chi_{R}(x)|\le R^{-2}B_{ji}\le M_{2}R^{-2} (claim 5 of Elementary Arithmetic in an Ordered Field). This proves claim 1. We record for Step 3 that if ∥x∥>2R\lVert x\rVert>2R then R−1x∈W2R^{-1}x\in W_{2}, so by Step 0

χR(x)=0,∂iχR(x)=0,∂j∂iχR(x)=0(∥x∥>2R).\chi_{R}(x)=0,\qquad \partial_{i}\chi_{R}(x)=0,\qquad \partial_{j}\partial_{i}\chi_{R}(x)=0\qquad(\lVert x\rVert>2R).

Step 2: the function s(x)=12∥x∥2s(x)=\tfrac{1}{2}\lVert x\rVert^{2}. By claim 2 of Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square, s(x)=12 dE(x,0Rq)2s(x)=\tfrac{1}{2}\,d_{E}(x,0_{\mathbb{R}^{q}})^{2}, so A Scaled Squared Distance to a Point is of Class C2C^2, with Gradient and Hessian applies with n=qn=q, U=RqU=\mathbb{R}^{q}, a=0Rqa=0_{\mathbb{R}^{q}}, c=12c=\tfrac{1}{2} and its function qq taken to be ss: by its claim 1, ∂js(x)=2⋅12 (xj−0)=xj\partial_{j}s(x)=2\cdot\tfrac{1}{2}\,(x_{j}-0)=x_{j}, that is ∂js=πj\partial_{j}s=\pi_{j}; by its claim 2, ss is of class C2C^{2}; and by its claim 3, D2s(x)=(2⋅12)Iq=IqD^{2}s(x)=(2\cdot\tfrac{1}{2})I_{q}=I_{q} (the matrix scalar multiple being that of Differential Calculus and Convexity on Euclidean Open Sets: Standing Notation (i)), so by The Hessian Matrix of a Function of Class C², ∂j∂is(x)=(Iq)ji=δji\partial_{j}\partial_{i}s(x)=(I_{q})_{ji}=\delta_{ji} for all i,j∈[q]i,j\in[q] and all xx. Moreover ss is smooth: x↦∥x∥2x\mapsto\lVert x\rVert^{2} is smooth by The Squared Euclidean Norm is Smooth, and ss is its scalar multiple by 12\tfrac{1}{2}, smooth by claim 3 of Constants, Affine Functions, Sums, Products and Quotients of Cᵏ Functions.

Step 3: claim 2. Fix R>0R>0. As ψR=χR s\psi_{R}=\chi_{R}\,s is a product of smooth functions, it is smooth by claim 3 of Constants, Affine Functions, Sums, Products and Quotients of Cᵏ Functions; it vanishes wherever χR\chi_{R} does, hence whenever ∥x∥≥2R\lVert x\rVert\ge2R, so it is compactly supported by claim 2 of Compact Support on Rn\mathbb{R}^n Means Vanishing Outside a Bounded Set. By the product rule of claim 1 of Constants, Affine Functions, Sums, Products and Quotients of Cᵏ Functions and Step 2, for all xx and i,j∈[q]i,j\in[q],

∂iψR(x)=∂iχR(x) s(x)+χR(x) xi,\partial_{i}\psi_{R}(x)=\partial_{i}\chi_{R}(x)\,s(x)+\chi_{R}(x)\,x_{i},

and, applying the sum and product rules of the same claim to this expression (whose two summands are products of functions of class C1C^{1}), with ∂jπi=∂j∂is=δji\partial_{j}\pi_{i}=\partial_{j}\partial_{i}s=\delta_{ji} by Step 2,

∂j∂iψR(x)=∂j∂iχR(x) s(x)+∂iχR(x) xj+∂jχR(x) xi+χR(x) δji.\partial_{j}\partial_{i}\psi_{R}(x)=\partial_{j}\partial_{i}\chi_{R}(x)\,s(x)+\partial_{i}\chi_{R}(x)\,x_{j}+\partial_{j}\chi_{R}(x)\,x_{i}+\chi_{R}(x)\,\delta_{ji}.

Put M=2M2+4M1+1M=2M_{2}+4M_{1}+1, which does not depend on RR; since M1,M2≥0M_{1},M_{2}\ge0 (Step 0) and 0≤10\le1 (claim 1 of Elementary Arithmetic in an Ordered Field), claim 2 there gives 0≤M0\le M and M1+1≤MM_{1}+1\le M, and 0≤∥x∥0\le\lVert x\rVert for every xx by claim 1 of Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square. Fix xx and i,ji,j. If ∥x∥>2R\lVert x\rVert>2R, all terms on the right of both displays vanish by Step 1, so ∂iψR(x)=0\partial_{i}\psi_{R}(x)=0 and ∂j∂iψR(x)=0\partial_{j}\partial_{i}\psi_{R}(x)=0, and the two bounds hold because their right sides are nonnegative (0≤M∥x∥0\le M\lVert x\rVert by claim 5 of Elementary Arithmetic in an Ordered Field). If ∥x∥≤2R\lVert x\rVert\le2R, then s(x)=12∥x∥∥x∥≤12 2R∥x∥=R∥x∥s(x)=\tfrac{1}{2}\lVert x\rVert\lVert x\rVert\le\tfrac{1}{2}\,2R\lVert x\rVert=R\lVert x\rVert and s(x)≤12(2R)2=2R2s(x)\le\tfrac{1}{2}(2R)^{2}=2R^{2} (claim 5 of Elementary Arithmetic in an Ordered Field and claim 2 of Monotonicity of Squaring on the Nonnegative Elements of an Ordered Field), ∣xi∣,∣xj∣≤∥x∥≤2R|x_{i}|,|x_{j}|\le\lVert x\rVert\le2R by claim 4 of Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square, and ∣χR(x)∣≤1|\chi_{R}(x)|\le1. Using claim 1, the multiplicativity and the triangle inequality of the absolute value (claims 4 and 5 of Properties of the Absolute Value in an Ordered Field) and claim 5 of Elementary Arithmetic in an Ordered Field,

∣∂iψR(x)∣≤M1R−1⋅R∥x∥+∥x∥=(M1+1)∥x∥≤M∥x∥,|\partial_{i}\psi_{R}(x)|\le M_{1}R^{-1}\cdot R\lVert x\rVert+\lVert x\rVert=(M_{1}+1)\lVert x\rVert\le M\lVert x\rVert, ∣∂j∂iψR(x)∣≤M2R−2⋅2R2+M1R−1⋅2R+M1R−1⋅2R+1=2M2+4M1+1=M.|\partial_{j}\partial_{i}\psi_{R}(x)|\le M_{2}R^{-2}\cdot2R^{2}+M_{1}R^{-1}\cdot2R+M_{1}R^{-1}\cdot2R+1=2M_{2}+4M_{1}+1=M .

Finally, ΔψR(x)=∑i=1q∂i∂iψR(x)\Delta\psi_{R}(x)=\sum_{i=1}^{q}\partial_{i}\partial_{i}\psi_{R}(x) satisfies ∣ΔψR(x)∣≤∑i=1q∣∂i∂iψR(x)∣|\Delta\psi_{R}(x)|\le\sum_{i=1}^{q}|\partial_{i}\partial_{i}\psi_{R}(x)| by claim 2 of Comparison and Absolute Value Bounds for Finite Sums of Real Numbers, and ∑i=1q∣∂i∂iψR(x)∣≤∑i=1qM\sum_{i=1}^{q}|\partial_{i}\partial_{i}\psi_{R}(x)|\le\sum_{i=1}^{q}M by claim 1 of the same lemma; and ∑i=1qM=M ι(q)=qM\sum_{i=1}^{q}M=M\,\iota(q)=qM by claim 3 of Properties of Finite Sums and The Canonical Map from the Natural Numbers to a Field. This proves claim 2.

Step 4: claim 3. Fix R>0R>0 and let B={y∈Rq:∥y∥<R}B=\{y\in\mathbb{R}^{q}:\lVert y\rVert<R\}. The set BB is open: if y∈By\in B and ∥y′−y∥<R−∥y∥\lVert y'-y\rVert<R-\lVert y\rVert, then ∥y′∥≤∥y∥+∥y′−y∥<R\lVert y'\rVert\le\lVert y\rVert+\lVert y'-y\rVert<R by claim 6 of Properties of the Euclidean Norm on Rn\mathbb{R}^n: Square, Distance, Vanishing, Coordinate Bounds, Homogeneity, Triangle Inequalities and Expansion of a Square applied to y′=y+(y′−y)y'=y+(y'-y), so the argument of Step 0 applies. By Step 1, χR(y)=1\chi_{R}(y)=1 for every y∈By\in B, so ψR(y)=s(y)\psi_{R}(y)=s(y) for every y∈By\in B; that is, ψR∣B=s∣B\psi_{R}|_{B}=s|_{B}. Let x∈Bx\in B and i,j∈[q]i,j\in[q]. By claim 1 of Restriction of a Cᵏ Function to an Open Subset applied to ψR\psi_{R} and to ss (both of class C1C^{1} on Rq\mathbb{R}^{q}),

∂iψR(x)=∂i(ψR∣B)(x)=∂i(s∣B)(x)=∂is(x)=xi,\partial_{i}\psi_{R}(x)=\partial_{i}(\psi_{R}|_{B})(x)=\partial_{i}(s|_{B})(x)=\partial_{i}s(x)=x_{i},

the last equality by Step 2. Since this holds at every point of BB, the functions ∂iψR\partial_{i}\psi_{R} and ∂is\partial_{i}s (both of class C1C^{1} on Rq\mathbb{R}^{q}, as ψR\psi_{R} and ss are of class C2C^{2}) have the same restriction to BB, and the same claim 1 gives ∂j∂iψR(x)=∂j∂is(x)=δji=δij\partial_{j}\partial_{i}\psi_{R}(x)=\partial_{j}\partial_{i}s(x)=\delta_{ji}=\delta_{ij}. Hence ΔψR(x)=∑i=1qδii=∑i=1q1=ι(q)=q\Delta\psi_{R}(x)=\sum_{i=1}^{q}\delta_{ii}=\sum_{i=1}^{q}1=\iota(q)=q by The Canonical Map from the Natural Numbers to a Field, and ψR(x)=s(x)=12∥x∥2\psi_{R}(x)=s(x)=\tfrac{1}{2}\lVert x\rVert^{2}. This proves claim 3.

Citations

Loading…

Dependencies

Uses0

Loading…

Comments

Log in to comment.

Loading…